mirror of
https://github.com/openai/codex.git
synced 2026-09-07 15:40:00 +00:00
38006a282e
## Summary - restore the existing `codex.tool_result` behavior and remove the added `codex.inference` event - emit `codex.tool_call` timing events only for direct tool calls - snapshot the execution-start marker once when emitting `codex.tool_call` so a concurrent dispatch start cannot produce internally inconsistent timing fields ## Why This keeps the stacked change focused on direct tool-call timing without expanding result or inference telemetry. The timing-marker fix addresses repeated reads of the marker during event emission, which could otherwise observe different states if dispatch begins concurrently. ## Validation - `just fmt` - `git diff --check HEAD~3..HEAD` - `just test -p codex-core tool_call_timing_guard_ignores_code_mode_source` - `just test -p codex-app-server app_server_emits_structured_tool_call_timing_event` - `just test -p codex-otel otel_export_routing_policy_routes_tool_result_log_and_trace_events` I also attempted `just test -p codex-core -p codex-otel -p codex-app-server`; that broader run was not completed after multiple app-server tests timed out.